#21 1952 · NRMP · Medical labour markets
American medicine built one clearing house and the offer wars stopped
the problem
Hospitals poached students with exploding offers
background
By the 1940s hospitals competed to lock in the best medical students earliest, pushing offer dates into the third year of school before students had seen enough of medicine to choose well. A 1945 embargo on releasing student records until an agreed date only compressed the problem: hospitals began issuing "exploding" offers with a reply deadline, which shrank from ten days in 1945 to under twelve hours by 1950 — students deciding over the phone with no time to compare alternatives.
A University of Chicago dean proposed the fix: instead of both sides racing to lock in a deal first, students and hospitals would each submit ranked preference lists to a central clearinghouse, which would compute the matches. In March 1952 the first Match paired 5,564 senior medical students with roughly 10,500 intern positions by an algorithm that, without anyone naming it yet, approximated what mathematicians would formalize nearly a decade later as deferred acceptance.
what everyone would do
The standard response to hospitals poaching students earlier and earlier was to constrain timing directly — the 1945 embargo on releasing student records until an agreed date — trying to slow the race down by regulating when offers could be made, while leaving the underlying decentralized, bilateral negotiation process itself unchanged.
what they saw
The Chicago dean saw that the timing embargo hadn't fixed anything, it had just compressed the same competitive pressure into exploding offers with shrinking reply deadlines. The real problem wasn't when offers happened, it was that the whole system relied on sequential bilateral negotiation under time pressure, where neither side could see the full landscape of alternatives before committing. Replacing that with a single, simultaneous, centralized clearinghouse where everyone submitted preferences once removed the race entirely, since there was no longer any advantage to committing early.
the move
The US residency Match: a clearing house running what was later proven to be the deferred-acceptance algorithm, redesigned by Alvin Roth in 1998.
why it works
Instead of each hospital racing to lock in students before rivals could, and each student deciding under pressure without full information, both sides submitted ranked preference lists to a central clearinghouse simultaneously — and because no offer or acceptance happened until all preferences were collected, there was no early-mover advantage left to exploit and no reason for either side to rush under an artificial deadline. An algorithm, later formalized as deferred acceptance, computed matches respecting both sides' stated preferences as well as the constraints allowed, replacing ad hoc bilateral negotiation with a systematic process where matches reflected actual fit rather than who negotiated fastest with the least information. The exploding-offer arms race collapsed entirely because its underlying incentive, locking in a candidate before a rival could, no longer existed once no individual party could act to secure a match before the clearinghouse computed the full result.
the payoff
Killed decades of exploding-offer chaos; still allocates every US resident today.
where it breaks
The mechanism requires participants to submit honest preference rankings for the algorithm to produce good outcomes, and even within a well-designed clearinghouse, who gets to 'propose' in the algorithm's mechanics has real distributional consequences — the case's own legacy notes the original hospital-proposing version quietly favored programs over applicants. It also struggled with a real edge case, dual-physician couples needing compatible matches across distant cities, that the basic algorithm wasn't designed to handle well, requiring the 1998 applicant-proposing redesign to address, showing centralized matching only solves the constraints it's explicitly built to weigh, not every constraint participants care about. And it requires enough volume on both sides to make a centralized process worthwhile — a market too small or too idiosyncratic in its preferences might benefit less from formal centralized matching than from direct negotiation.
what came after
The original hospital-proposing algorithm quietly favoured programs over applicants and struggled with dual-physician couples, whom it often split across distant cities — enough of a problem that some couples began negotiating around the Match entirely. The NRMP commissioned Alvin Roth in 1995 to redesign it; his applicant-proposing algorithm with Elliott Peranson, adopted in 1998, fixed the couples problem and shifted the balance toward applicants, though later analysis found the change affected well under 1% of individual matches. The Match still runs every year: in 2022 it placed the large majority of roughly 42,500 applicants into about 36,300 first-year residency positions.
references
- [1]Match Day 101: How does the medical residency match work?Stanford Medicine, 2024med.stanford.edu
- [2]The Effects of the Change in the NRMP Matching AlgorithmJAMA (via PubMed), 1997pubmed.ncbi.nlm.nih.gov